Piano World Home Page Forums Our Most Popular Forums Digital Pianos - Electronic Pianos - Synths & Keyboards Sampled VS Modelled Pianos ???

I think for the most part those musicians are ill informed or being paid by Roland or PT. It doesn't take a lot of effort to try both and find that the V and especially PianoTeq are much less realistic and beautiful to listen to than a quality sampled piano.

Sampled pianos already use modelling to add stuff like sympathetic resonance. Synthogy calls it:



Since modelled pianos all seem to use some kind of base tone as well that stays the same no matter the preset, I really don't see why this "samples are just a snapshot" argument comes up so often. Listen to those Pianoteq D4 preset comparisons:

http://www.pianoteq.com/d4_in_action?s=d4_presets

They all have the same base tone with just slight variation in equalization, softness, reverb etc. but the general sound stays the same.

Roland admitted to the same thing, they had to use some kind of base tone in the V-Piano that remains no matter what. This isn't really suprising. In synthesis, there is always some kind of base tone involved, usually in the form of oscillators. All that you call "modelling" is what follows after that to mold the sound into something more interesting.

And whoever claims that P4 sounds better than any sampled piano, I do hope you're kidding. I really do. You must have an unbelievable grudge against sampled pianos to claim such a thing. Pianoteq 4 is certainly an improvement over 3, but listen to these demos, then listen to some Ivory II demos, then listen to some real piano recording. One sounds like a toy, two sound like the real deal. Surely you can figure out which is which. If not then you have lost either objectivity or hearing.

Excellent post, Gigantoad.



I think it's the perpetuation of some marketing, honestly. The best sampled pianos use recordings to pin down the piano sound at a few velocity points and then interpolate between to get a sound for each of the 127 velocities. Pure modeled pianos just make up a sound for each velocity, but in both cases for a given velocity the sound is fixed. There's no reason to think one sounds more dead than the other.

For that matter, talk of PT being more "playable" is the perpetuation of marketing as well, I think. Back in the day sampled libraries didn't have partial pedaling or repedaling, and they didn't have nuances like sympathetic resonance and release samples. That got labeled playability. Somehow PT persuaded people that the lack of those items is inherent in having a sampled base sound (timbre). But it's not. PT and sampled libraries both must program in these features explicitly. Modeling refers only to the generation of the underlying timbre. In short, I do not believe that sampled pianos are more playable. None that I have played are more playable or natural-responding than Galaxy, for example. Even if you can find some way in which a sampled piano does not behave right, it just means the sampled piano maker didn't think to add programming for that case. It has nothing to do with the engine that generates the timbre.

Attempts to model piano are similar to attempts to have computers speak and make it sound natural. Scientists have been working on this since computers were in their infancy, and they have improved a lot over time. But it still doesn't sound right. Making an algorithm generate a sound (from scratch) that is generated by a nontrivial process in real life is hard. I expect that modeled pianos will sound better than sampled pianos when modeled voices sound better and more natural than recorded human voices. That day is far off and it may never come.

Modeled pianos can be very tweakable, and the technology is really cool, but that doesn't mean in their current state they compare with serious sampled libraries.

not good comparison because our eyes are far more easily fooled then ears, that's the reason.

Hmmm. That doesn't sound generally true to me. I don't see eyes being more easily fooled than ears in general. Hard things to compare, but I think eyes are relatively difficult to fool. My wife can't tell the difference between a recording of a cheapo keyboard, a nice digital, a VST, and an acoustic. She can tell very easily that the robot in Terminator 1 is stop-motion and fake.

I think alekkh is right. Modeled sound could progress much faster if

1. There was a ton more money behind it.

2. There was a compelling need for it. Hollywood only uses computer special effects where it's too expensive or difficult/impossible to do with real life stuff. Otherwise they use real life stuff.

I don't see a ton of money flowing to piano sounds any time soon, and the need to move away from sampled technology doesn't strike me as very compelling. That's why there are only two fully modeled pianos in the world and they aren't as good as they could be.

you got to be kidding, I suggest go back to high school physics.
This is well known fact, as you know you can fool you eyes simply
by sorting fast pages of a book hence we have a movies today.
Try this with ears where some fellows can distinguish fraction of half tone differences.



p.s.
My bad BTW getting into arguments which piano sound is better. It's very subjective plus people are very intolerant of others having drastically different opinions. I know what I know and what I hear and you know what you hear (or not) - all is an illusion anyway. Also FYI The reason I think sampled pianos will go away is the same as with vinyls faded away.

Eyes are certainly hard to fool, especially when we see simulated things that we are used to seeing everyday like faces. It took the 3D industry very long to create believable 3D faces with correct animations. On the other hand, sound engineers have used all kinds of materials to simulate sounds in movies for decades, like horse galloping simulated with coconuts.

Granted, as gvfarns has pointed out, simulating voices is very hard as well because we're so used to how they sound and we can make out even the smallest inconsistency. Maybe that comparison fairer in terms of piano modelling.

There is another aspect which I would call "habit". When you listen to something or watch something long enough, you will get used to it. We can observe this phenomenon when people buy new TV's. The colors might be completely off but after just a few days they will get used to it. At some point they might even be convinced that their TV has the most realistic colors on the market. Only when their sets are properly color-calibrated will they realise, slowly, how very wrong the colors were before.

I think the same thing is happening with Pianoteq. There is a minority, most prominent in the PQ forum obviously, with the impression that Pianoteq 3 was the holy grail in term of piano sound reproduction sounding much more like a real piano than anything else. Now comes Pianoteq 4 and pretty much everyone over their calls it a major upgrade (which it is imo). Doesn't that show that PT3 was not anywhere near as good as they thought? And now the circle of denial starts again.

Wrong again. Many shops are giving over more floor space to vinyl records now - there is a real resurgence in interest in vinyl (I'm talking new releases, not second hand records). There may be lots of reasons for this but I can tell you one may be that they simply sound better.

+1

check this one out and try this with ears...



or this, which direction the girl is spinning? because it can be either way...

Oh no, a return to vinyl because it sounds better.

When CDs were introduced many folks thought they sounded sterile. The reason was, the vinyl LPs were mastered from tapes with their inherent tape hiss. When that hiss was not present we missed it and things sounded sterile. (Also some folks just have an aversion to anything digital, it has be analogue to be good.)

I guess now's the time for a discussion for those extremely expensive audio cables, you know, the cables that cost more per meter than most of us make in a week.

Now is not that time!

http://listverse.com/2008/02/29/top-10-incredible-sound-illusions/

Either way, optical illusions are no proof that our eyes are easily fooled. They merely exploit certain ways how our brain processes images. Using this to back your argument is actually pretty dumb because neither images nor sound are usually aimed at creating such illusions. When you watch a movie you will never be tricked in such ways, but you will very easily notice if that 3D face isn't entirely realistic compared to that of a real actor.

Oh, for the love of god, please don't start up that debate again...

piano_shark - now you are in trouble. No number of images or animations posted will save you. That's a statement that demands a scientific proof. Which research has shown that? Only credible sources will count.


I think you wouldn't find one. Because your are comparing apples with carrots.

OP here

Again thanks to all for your contributed opinions it's been a long an debated discussion. However the outcome of my choice is to go with a Sampled Library, that being Galaxy 2, Galaxy Vintage D and the amazing sound from Imperfect Samples.

I think going with these library's after certain FXs have been applied in the creation process, Tone, Resonance, Articulation, Convolution Reverb etc, and of course a good Master, i would hope after post production the recorded Material should be at a standard acceptable for retail.

A modeled piano can create 127 distinct sounds for those 127 velocities, each containing numerous parameters that are gradually and smoothly varied from steps 1 through 127. Unless a sampled piano includes 127 sampled velocity layers, the equivalent interpolations will be "bumpier" -- i.e. cannot be entirely smoothly gradated from point 1 to 127 because the underlying sound being manipulated has distinct shifts at various points (even though they may try to minimize the audible effects with layer blending). Also samples cannot be modified in all the same ways, because post-processing doesn't have access to all the same elements. For example, if you want to manipulate just the third harmonic of the tone in a particular way over time as velocity changes, you can easily extract and manipulate that in a model, not really in a sample.

I think these are the kinds of things that allow the dynamic gradations of a model to generally feel more sophisticated and more seamless than the velocity gradations in sample based instruments, which I think by nature are more coarse and primitive at their core.

Also, AFAIK, no sample based instruments provide for more than 127 variations. The Pianoteq model provides for more (and the VAX77 controller can trigger them).

OTOH, there are qualities that modeling hasn't successfully captured this far, such that a sample at a given velocity can sound more authentic than a model at the same velocity. So I think that's why we come to the perspective that, at least as a generality, the sampled sound is more accurate, but the playing transitions from one velocity to another is smoother in models. So each can be seen as more realistic than the other in some aspect.

The slight issue is that those transitions are perhaps noticed by some, but not by many. I would hazard a guess that most people do not hear the difference between 127 velocity levels and 18 as long as the latter has smooth layer blending going on. On the other hand, if the sound as a whole sounds as unconvincing as Pianoteq, then that is much more noticable.

exactly - playability, dynamics and control are much more important for serious player then just simply "nice sound".
Most of defending here sampled piano are kids raised on digital
instruments to start with so no wonder the low expectation...
Yamaha and other companies know that that's why still selling sampled pianos for masses.